v0.14.2

What's Changed

• Release v0.14.2: moviepy skill, Python examples, dep cleanup (66a2c89)
• FIX: cross-platform font loading in moviepy examples (3057ab8)
• DEPS: declare moviepy/Pillow/matplotlib in tools/requirements.txt (6f94523)
• EXAMPLE: add hello-world — the quick-start the README already promised (2a638fe)
• EXAMPLES: add quick-spot and data-viz-chart moviepy references (4402e5a)
• SKILLS: add moviepy skill, audio-anchored timelines, LTX-2 cameo pattern (9a57fdf)
• README: add q2 townhall longarm-ad and stars to examples (c6df139)

Full Changelog: v0.14.1...v0.14.2

v0.14.1

What's Changed

• Release v0.14.1: chain_video, acemusic default, progress reporting (ffe9929)
• Harden chain_video: timeout, error handling, cleanup, and registry entry (a076f87)
• SKILL: promote yieldMs polling to top-level CRITICAL rule (84a04c1)
• SKILL: add style drift prevention rules for chained LTX-2 (c99aa08)
• Fix chain_video: glob for suffix variants when finding previous clips (65f9f4a)
• SKILL: add yieldMs + process poll pattern for live progress (501c71d)
• Add chain_video.py: sequential LTX-2 with visual continuity (c3218f9)
• Merge remote-tracking branch 'origin/main' (cde5f43)
• Add structured progress reporting for all cloud GPU tools (e18aab5)
• Add acemusic cloud API as default music generation provider (b9f3956)
• Sync official Remotion skills (upstream d5d3955) (74324e7)
• Add R2 recovery note to SKILL.md for SadTalker client timeouts (b5573ac)
• Increase SadTalker timeouts: 20x multiplier + 300s buffer for size=512+gfpgan on A10G (42a54de)
• Release v0.13.3: SadTalker dynamic timeouts and pre-flight estimates (e3ea1bb)
• Release v0.13.2: Default to Modal, lint fixes (51a3e00)
• Remove unused imports in ltx2.py (json, requests) (8862852)
• Default all cloud GPU tools to Modal instead of RunPod (977a397)
• Bump version to 0.13.1 (e0d8dd7)
• Update docs, README, registry, and changelog for LTX-2 (v0.13.1) (2710053)
• Add LTX-2 skill and update openclaw skill with video generation (87ed96f)
• Add LTX-2.3 video generation via Modal (text-to-video, image-to-video) (d26b043)
• Harden OpenClaw skill for autonomous use (add6661)
• Add full setup flow to OpenClaw skill (b3ebe5a)
• Add OpenClaw skill for autonomous video creation (alpha) (6a7ee1c)
• Add OffthreadVideo guidance: never use raw
• Fix lib/transitions barrel: remove @remotion/transitions re-exports (98f3344)
• Fix R2 region, SadTalker timeouts, and tool path guidance (ce63b78)
• Clarify Modal free tier capacity: a few 5-minute videos per month (f9d5d62)
• Reframe 'what's free' around open-source models, not just pricing (23ea71e)
• Fix Cloud GPU section: Modal builds its own images, not GHCR (fb51478)
• Rewrite modal-setup.md: Modal as recommended provider, add R2 and /setup refs (11f9844)
• Restructure README: quick start first, eliminate overlap (0552039)
• Add /setup wizard, verify script, and migrate dewatermark to cloud_gpu (eb90d47)
• Add Modal docs, propainter app, and cloud GPU quick start (2eec07a)
• Track Modal cloud GPU integration in ROADMAP.md (54e6b93)
• Fix sadtalker Modal app: numpy-torch bridge and basicsr compat (e0f6cab)
• Fix Modal app build issues discovered during deployment (258a929)
• Add Modal apps for upscale, image-edit, music-gen, and sadtalker (Phase 3) (a7f7f8e)
• Migrate all cloud GPU tools to shared abstraction (Phase 2) (5bebb53)
• Add Modal as alternative cloud GPU provider (Phase 1) (303c692)

Full Changelog: v0.14.0...v0.14.1

v0.14.0

What's Changed

• Add acemusic cloud API as default music generation provider (09a87a6)
• Merge pull request #7 from digitalsamba/auto/sync-remotion-skills (0437fef)
• Sync official Remotion skills (upstream d5d3955) (19076e1)
• Add R2 recovery note to SKILL.md for SadTalker client timeouts (d911464)
• Increase SadTalker timeouts: 20x multiplier + 300s buffer for size=512+gfpgan on A10G (9250110)
• Release v0.13.3: SadTalker dynamic timeouts and pre-flight estimates (7a69eec)

Full Changelog: v0.13.2...v0.14.0

v0.13.2

What's Changed

• Release v0.13.2: Default to Modal, lint fixes (e0e2d88)
• Remove unused imports in ltx2.py (json, requests) (8289f4f)
• Default all cloud GPU tools to Modal instead of RunPod (b98933c)

Full Changelog: v0.13.1...v0.13.2

v0.13.1

What's Changed

• Merge pull request #9 from digitalsamba/feature/ltx2-integration (6e65782)
• Bump version to 0.13.1 (c03f050)
• Update docs, README, registry, and changelog for LTX-2 (v0.13.1) (ab50c25)
• Add LTX-2 skill and update openclaw skill with video generation (d3a33cd)
• Add LTX-2.3 video generation via Modal (text-to-video, image-to-video) (cad056a)

Full Changelog: v0.13.0...v0.13.1

v0.13.0 — Modal Cloud GPU + /setup Wizard

What's New

Modal as recommended cloud GPU provider — All 7 AI tools now support --cloud modal alongside --cloud runpod. Modal offers faster cold starts, simpler deployment, and $30/month free compute on the Starter plan.

/setup wizard — Interactive first-time setup that walks through cloud GPU deployment, Cloudflare R2 file transfer, and voice configuration. Replaces the manual per-tool --setup flow.

Shared cloud GPU abstraction (tools/cloud_gpu.py) — Unified interface for all cloud tools. Eliminates ~2,000 lines of duplicated RunPod boilerplate across 7 tools.

OpenClaw skill (alpha) — Autonomous video creation skill for creating videos with minimal human intervention.

Migration

Existing RunPod setups continue to work unchanged. All tools still accept --cloud runpod. No .env changes required.

To add Modal alongside RunPod:

pip install modal
python3 -m modal setup
# Then in Claude Code:
/setup

Other Changes

• 7 Modal apps in docker/modal-*/app.py — Modal builds containers from these, no GHCR needed
• tools/file_transfer.py — Shared R2/fallback upload module
• tools/verify_setup.py — Health check for all configured endpoints
• Fixed lib/transitions barrel re-exporting invalid @remotion/transitions entries
• Added CLAUDE.md guidance: always use <OffthreadVideo>, never raw <video> in Remotion
• README restructured: quick start first, per-tool cost estimates, Modal recommended
• docs/modal-setup.md — Full Modal setup guide

Cost Estimates (Modal)

Tool	Per Job
qwen3_tts	~$0.01
flux2	~$0.02
image_edit	~$0.03
upscale	~$0.01
music_gen	~$0.05
sadtalker	~$0.10
dewatermark	~$0.10

Typical monthly spend: $1-2. Well within Modal's free tier.

v0.12.0

What's Changed

• Merge feature/acestep-music-gen: Add ACE-Step 1.5 music generation (v0.12.0) (cc90f37)
• Update docs, README, and changelog for v0.12.0 release (3092bd8)
• Add ACE-Step skill with prompt engineering and video production patterns (7fbe2d4)
• Fix gen_params reference in RunPod handler cleanup (cfd5838)
• Add ACE-Step 1.5 music generation tool with RunPod deployment (929dc74)

Full Changelog: v0.11.1...v0.12.0

v0.11.1

What's Changed

• Add RunPod queue timeout safeguards and R2 upload fixes (v0.11.1) (6060c43)

Full Changelog: v0.11.0...v0.11.1

v0.11.0

What's Changed

• Add FLUX.2 Klein 4B image generation tool with scene presets (v0.11.0) (079a3a7)
• Merge pull request #5 from digitalsamba/auto/sync-remotion-skills (3bdad4d)
• Sync official Remotion skills (upstream 937ffb3) (ab21f47)
• Add v0.10.1 changelog entry (31c34b1)

Full Changelog: v0.10.1...v0.11.0

v0.10.1

What's New

New Tool: sync_timing.py — Audio-to-Config Timing Sync

After generating per-scene voiceover audio, TTS durations often differ from the estimated durationSeconds in your Remotion config. This tool automates the feedback loop:

python3 tools/sync_timing.py              # Compare actual vs config
python3 tools/sync_timing.py --apply      # Update config with 1s padding
python3 tools/sync_timing.py --json       # Machine-readable output

Features:

• Auto-detects config file and template type (sprint-review v1/v2, product-demo)
• 3-pass audio-to-scene matching: audioFile field → index → name
• Uses math.ceil(audio + padding) for whole-second durations
• Skips changes < 0.3s to avoid churn
• Creates .bak backup before modifying config
• Accepts voiceover.py --json output directly via --voiceover-json
• Suggests playbackRate adjustments for demo scenes

CLAUDE.md Slimmed Down (44% smaller)

Reduced from 861 → 480 lines by removing catalog data duplicated in toolkit-registry.json and adding cross-references. All workflow guidance, timing knowledge, code patterns, and tool-specific gotchas retained.

Removed (now in registry only): skills table, commands table, components table, transitions table, preset lists, Docker image names, duplicate CLI examples.

Kept: production workflow, project lifecycle, video timing (pacing tiers, word budgeting, TTS drift patterns), Remotion code patterns, transition code examples, design refinement guidance, SadTalker/NarratorPiP critical flags, redub sync explanation.

Full Changelog

v0.10.0...v0.10.1